42
Chapter
5. 0.18 /Lxm CMOSTechnology
Process Substrate Metals
Supply Voltage
l'ox
p-type/Triple
Well6
layers
1.8 V
Chapter 6
Design of the Receiver
6.1 Nonlinear Distortion of MOSFETs
In this
section,
weestimatethemagnitude
of the intermodulationprod¬
ucts
generated by
the nonlinear I-V characteristic ofa MOS transistor.In
particular
wecalculate theinput
referred third orderintercept point
of a common sourceamplifier assuming
that the I-V characteristic of the transistor is the dominantnonlinearity.
The results areuseful as ayardstick during
the firstphase
of thedesign.
We describe the characteristic of the device as
[10]
Id(vGs)
= oh , fl,ß
tttt(«gs-Vp)2 (6.1)
2[1 A-0(vGs -
Vt)\
where
ß
isthe transconductancefactor,
and 6 isamobility degradation
factor which for the usedprocessis around 3
V~1.
Since the intermod¬ulation
specifications
describe low levelnonlinearities,
weexpand
theabove I-V characteristic in a
Taylor
series and truncate it after thethird order component:
dld
Ä 1d2Id
2 1^.3-r àvas A-
t^1Tö-övgs
+ ^7T^Th' dvGS 2! dv2GS L"s 3!
dvGS Id(vGS)
«Ido
+ ^r—SvGS + 7-1-^Sv2GS + -—r-Sv=
Id0
A-a-iSvas
A--y5v2GS
+-^Svas (6-2)
When two
signals
ofequal amplitude
areapplied
to theinput,
i.e.Svqs
=A[sm(u>it)
A-sm(u>2t)],
the third ordercomponent of theTaylor
43
44
Chapter
6.Design
of the ReceiverFigure
6.1:Input referred
third orderintercept pointof
a MOS tran¬sistor.
expansion
generates thefollowing
sidebands:a3
~3\°Vgs-5v3r as
A*
6
'
4
{3sin[(wi -2w2)t]
+3sin[(2wi -w2)t]}
+...(6.3)
The third order
intercept point
isdefinedasthesignal
levelatwhichthe above sidebands have the same
amplitude
astheinput signals
am¬plified by
the first order term a\Svgs
and can thus be calculatedas:
CL\ A =
=> A =
Hi .A3-
Inserting
the values ofa\ and as, calculated from(6
])
we obtain:A — A
V2
2
vod(i
+evody(2
+evod)
3
1/
0where vod is the overdrive
voltage
defined asVod =
VGso
-Vp
(6.4) (6.5)
(6.6)
(6.7)
6.2. Low Noise
Amplifier
45V,nO
Figure
6.2: LNA input stageconfigurations.
and
Vgso
is the gate-source biasvoltage.
We see that the iIP3 is afunction ofvod. The
higher
the overdrivevoltage,
thehigher
the iIP3(Fig. 6.1).
At overdrivevoltages
lower than about 100mV,
the transis¬tor enters the moderate inversion part of the I-V characteristic which deviates
substantially
from(6.1 ).
At overdrivevoltages
lower than 0 V the transistor is in weak inversion and its characteristic becomes expo¬nential
[31].
As aresult,
the iIP3 for vod < 100 mVdegrades rapidly approaching
avalue of—12.7+201og(n)
dBu fornegative values,
wheren is the
slope
factor which for the used process is 1.43.6.2 Low Noise Amplifier
The characteristics of the LNA to be realized are shown in
Fig.
4.5.In
addition,
toproperly
terminate thepreselection filter,
a 50 ilinput impedance
isrequired.
The 12 dB ofgain
can beimplemented
witha
single
stageamplifier, and,
to avoid the need for abalun,
we use asingle-ended topology.
For this
design
the threeinput
stageconfigurations depicted
inFig.
6.2 have been considered:configuration (a)
makes use of feed¬back torealizeabroadband
input matching.
Due tofeedback thestage is very linear andrelatively
insensitive toparasitics,
but torealize therequired input impedance
andgain
itrequires
alarge
transconductance and thus alarge
current.Configuration (b)
is a common-gate stage.The
input impedance
is definedby
the transconductance gm of the transistor.By making
gm = 20 mS a broadband 50 ilinput imped-
46
Chapter
6.Design
of the ReceiverFigure
6.3:Simplified
schematicof
the LNA.ance is realized. The
linearity
of the stage is poorer than that of(a),
but it is still much better than
required
for thisapplication (Fig.
61).
The minimum achievable noise
figure
is 2.2dB.1 Configuration (c)
is acommon-source stage in which the gate-source
capacitance Cgs
of thetransistor is embedded in a
matching
network totransform the capac¬itive
input impedance
of the transistor to 50 il. ForQ
greater than one, thematching
network formedby Ls, Lg,
andCgs
alsoprovides
some
voltage gain
between theinput
and the gate-source terminals of the transistor so that the effective transconductance of the stageGm
=1-o/Vm
islarger
than the transconductance gm of the transis¬tor.
Therefore,
toachieve thespecified gain,
alower currentisrequired
than with acommon-gate
configuration.
On thenegative side,
the volt¬age
gain
of thematching
networkdegrades
somewhat thelinearity
ofthe stage, but not to the
point
topreclude
its use for theapplication
at hand. Due to the
tight
constraint on powerconsumption
we have decidedto useconfiguration (c).
Figure
6 3 shows the schematic of thecomplete
LNA(without
bi¬asing)
. Ananalysis
of the circuit withanidealized transistor model in¬cluding only
the transconductance gm and the gate-sourcecapacitance Cgs provides
thefollowing design equations
for theinput impedance Zm,
thevoltage gain Av,
and the noise factorFld
calculated consider-1Assuming 7=2/3
6.2. Low Noise
Amplifier
47ing only
the drain noise current source of MZin
=ju(La
+Lg)
+-^-
A-^ (6.8)
Av — Gm Zip
=
-JQ
9mZL Zp
JUqLs (6.9)
Fid
=l+7n2
p(6.10)
Here
Rs
is the source resistance andZp
is the total loadimpedance
which is the
parallel
connection ofRd, Ld,
theoutputimpedance
of thecascodestage, and the
capacitive
loadCp provided by
the mixer drivenby
theLNA. The inductorLd
isused to tune-outCp
and theparasitic
drain-bulkcapacitance
ofM2
sothat for thedesign only
the real part of the loadimpedance
has to be considered. ResistorRd
is used to lower thequality
factorQp
of the load resonator and toprecisely
setthe
gain
of theamplifier.
Cascode transistorM2
is usedtoimprove
theunilaterality
of theamplifier.
One of the functions of theamplifier
is infact to minimize the emission of
spurious signals coming
from the LOthrough
the mixer.From
(6.9)
and(6.10)
it is evident that ahigh Q simultaneously
reduces the noise factor and the
required
gm. On the otherhand,
ahigh Q
also reduces the bandwidth of thematching network, and,
if madetoohigh,
calibration becomesnecessary. As thematching
network hastobeintegrated along
with theamplifier,
tolimit theimpact
ofcomponenttolerances,
thequality
factorQ
hastobe limitedtovalues in therange of 2-3. From(6.9)
it is also apparent that the transconductance of theinput
stageGm
is setby Ls.
Ahigh Gm requires
a smallLs
whichcanreadily
be realized withashortbonding
wire.Lg
andLd
areimplemented
ason-chip spiral
inductors. The values of thecomponents used toimplement
our prototype arelisted in Table 6.1.The current
required by
theamplifier
is 1.2 mA and is setby
lin¬earity requirements.
TransistorM\
is in fact biased at theboundary
48
Chapter
6.Design
of the Receiverit, _L,
GMUM2
Lg
7.5 nH W 104x2.5 iiiaLd
6.1 nH L 0.18 iiiaLs
1.8 nH 9m 18 mSRd
250 ilh
1.2 mAcc
3.7pF
Table 6.1: LNA
Design
Values.between strong and moderate inversion
(Vgs —Vp
« 80mV).
A fur¬ther reduction of thecurrent
(without
achange
in transistorgeometry)
would have made the
linearity
of thesystemdependent
uponmoderate inversion characteristics which are not well modeled and thus difficult topredict
and to control. A simultaneous reduction of current and transistor width in such a way as tokeep
the transistor in strong in¬version would have